NBER WORKING PAPER SERIES

THE SMUGGLING OF ART, AND THE ART OF SMUGGLING:UNCOVERING THE ILLICIT TRADE IN CULTURAL PROPERTY AND ANTIQUES

Raymond FismanShang-Jin Wei

Working Paper 13446http://www.nber.org/papers/w13446

NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts Avenue

Cambridge, MA 02138September 2007

We thank Daron Acemoglu, Ben Olken, Zhi Wang, and particularly Patty Gerstenblith for very helpfuldiscussions, and Andre Heng and Chang Hong for superb research assistance. The views expressedherein are those of the author(s) and do not necessarily reflect the views of the National Bureau ofEconomic Research.

© 2007 by Raymond Fisman and Shang-Jin Wei. All rights reserved. Short sections of text, not toexceed two paragraphs, may be quoted without explicit permission provided that full credit, including© notice, is given to the source.

The Smuggling of Art, and the Art of Smuggling: Uncovering the Illicit Trade in CulturalProperty and AntiquesRaymond Fisman and Shang-Jin WeiNBER Working Paper No. 13446September 2007JEL No. F1,K42,O1,Z11

ABSTRACT

We empirically analyze the illicit trade in cultural property and antiques, taking advantage of differentreporting incentives between source and destination countries. We thus generate a measure of illicittrafficking in these goods based on the difference between imports recorded in United States' customsdata and the (purportedly identical) trade as recorded by customs authorities in exporting countries.We find that this reporting gap is highly correlated with the corruption level of the exporting countryas measured by commonly used survey-based indicies, and that this correlation is stronger for artifact-richcountries. As a placebo test, we do not observe any such pattern for U.S. imports of toys from thesesame exporters. We report similar results for four other Western country markets. Our analysis providesa useful framework for studying trade in illicit goods. Further, our results provide empirical confirmationthat survey-based corruption indicies are informative, as they are correlated with an objective measureof illicit activity.

Raymond FismanGraduate School of BusinessColumbia University622 Uris Hall3022 BroadwayNew York, NY 10027and NBERrf250@columbia.edu

Shang-Jin WeiGraduate School of BusinessColumbia UniversityUris Hall, Room 6193022 BroadwayNew York, NY 10027-6902and NBERshangjin.wei@columbia.edu

2

1. Introduction

The smuggling of antiques and cultural property is thought to be big business. All countries

impose restrictions on the export of various classes of cultural property and antiques,1 ranging

from archeological objects to coins to older art works.2 Hence their sale abroad often requires their

illegal export from the country of origin. As with other activities of questionable legality, however,

it has been difficult to put a precise figure on the full extent of trafficking in cultural goods. For

trade in antiquities (unearthed ancient objects), which makes up only one component of the total

illegal trade in cultural objects, estimates ranging from $300 million up to $6 billion per year

(Atwood, 2004). According to Interpol’s estimates, the antiquities trade on its own ranks behind

only drugs and arms in its scale of illegal trafficking (Toner, 1999). Collectively, these illicit

activities represent the darker side of globalization – smuggling requires extra-legal activities that

may abet corruption, impose a strain on international relations, and potentially dampen the gains

from legitimate international trade.3 Thus, illicit trade is an important element of political economy

and international trade. Unfortunately, we have little systematic knowledge of the dynamics of

illicit trade, as data on illegal activities are by their very nature difficult to obtain.

In this paper, we analyze the illicit trade in cultural objects by taking advantage of a unique

aspect of their trade relative to other forms of smuggling: The stark difference in the legality and

legal enforcement of a particular shipment between exporting and importing countries. In

particular, the exportation of broad classes of cultural objects is prohibited by most countries

without a special permit. However, once these (illegally) exported goods have left the country of

origin, they are not generally regarded as contraband when imported into their destination, absent

additional agreements that we discuss below (Gerstenblith, 2008). In the United States specifically,

there is actually a strong incentive to report accurately on the importation of cultural objects: Any

goods entering the United States that are not properly declared are subject to customs seizure;

further, the zero tariff rate on antiques and cultural objects entering the country removes any

incentive to misdeclare valuation (U.S. Department of Homeland Security, 2006). Even in cases 1 Henceforth referred to simply as cultural objects or antiques. Throughout this paper we will be considering those products that, by international trade classification, belong to Harmonized System (HS) Product Code 9706 – Antiques of an age exceeding one hundred years. 2 The specific classes of objects that are restricted from export as well as the rules for gaining permission to export restricted objects differ across countries. The rules defy simple categorization or measurement of restrictiveness. See Prott and O’Keefe (1988) for the most recent comprehensive description of these laws worldwide. 3 See Andreas (1998) for an overview of these issues.

3

where importation is of questionable legality, differences in the burden of proof between exporting

countries and the U.S. generally allow for the relatively easy import of goods whose export would

not have been permitted by the source country.

As a result of these asymmetric reporting incentives, reported imports of cultural objects

into the United States provide a plausible measure of the ‘true’ level of trade in these goods that

we may compare with the export levels reported by cultural object-rich countries. The difference

between these two trade figures provides a credible measure of illegal exports.

What allows for the illicit export of cultural objects from the source country? Not

surprisingly, when smugglers are apprehended and their operations exposed, their activities are

often found to be facilitated through the bribing of customs officials to look the other way (Brody

et al, 2000). Hence, the illegal and unreported export of cultural objects is relatively easy in

countries with corrupt bureaucracies that allow for this type of transaction. Hence, if cross-country

survey-based measures of corruption do indeed reflect underlying corruption realities, these

measures should be good predictors of patterns of global trafficking in cultural objects. In this

sense, we may use our measure that is derived from objectively measured trade data to assess the

validity of these corruption indices that are often based on subjective perceptions.

In this paper, we present an objective measure of smuggling in cultural objects based on

this reporting gap between recorded exports on an exporter’s side and the recorded imports by U.S.

Customs. Without smuggling (and measurement error), the reporting gap should be zero. If the gap

were pure measurement error, it should not be correlated with country-level attributes. However,

we find that our smuggling measure is very highly correlated (with correlation coefficient =0.52)

with standard cross-country survey-based corruption indices, thus providing compelling and

objective validation of these indices. This pattern is robust to the inclusion of region effects and

controls for countries’ endowment of desirable/collectible cultural objects. Interestingly, our

smuggling variable is uncorrelated with the log of income per capita once the exporter’s corruption

level is controlled for, so it is unlikely that we are simply picking up the effects of country-level

wealth.

Several additional tests lend further support to our interpretation of the results. First, the

corruption-smuggling gap relationship is stronger for object-rich countries. Second, we run a

placebo regression using data on the reporting gap in the U.S imports of toys between the

exporter’s and US customs (U.S. reported imports of toys from a country, minus that country’s

4

reported exports of toys to the U.S. in the same year). Similar to imports of cultural objects into the

United States, toy imports also have a zero tariff rate on the U.S. side. In this case, we observe no

correlation between an exporter’s corruption level and the customs reporting gap, suggesting that

cultural objects do indeed present a special case. Finally, we report results for four other countries

– Canada, Germany, Great Britain, and Switzerland – all with zero tariffs on cultural objects that

are also reported to have a significant trade in these goods; we find a positive relationship between

corruption and the ‘smuggling gap’ for all four countries.

Our paper thus makes two primary contributions: Most importantly, we provide a first

empirical analysis of the trade in restricted goods, and further provide comparable cross-country

estimates on the smuggling of contraband. 4 We thus contribute to the growing literature on

measuring underground activities using differential reporting incentives (see, for example, Fisman

and Wei, 2004; Yang, 2007; Mishra et al, 2007). However, we highlight two key departures from

the prior literature: First, earlier studies have largely focused on a single exporting country;

second, previous research has focused on tariff evasion rather than the trafficking of illegal objects.

By contrast, cultural objects imports face no tariffs in the U.S. and other major markets (hence

tariff evasion is not the motivation) but are often subject to export controls in the country of origin.

Second, we provide a clear validation of subjective corruption indices based on objective trade

data.

The rest of this paper is organized as follows: Section 2 provides a short background on

laws governing the trade in antiques and cultural goods. Section 3 provides a description of the

data, and Section 4 presents our results. Section 5 concludes.

2. Legal background on international trade in cultural property and antiques5

Goods that have been illegally exported from one country are not generally regarded as contraband

when imported into the United States, absent some further agreement.6 In the case of cultural

4 Another related paper is Fisman and Miguel (2006), who using parking violations of U.N. diplomats in New York as a cross-country measure of corruption norms. Relative to that study, our method has the advantage of focusing on customs, a much larger (and often notoriously corrupt) branch of the civil service in many countries. 5 This section draws heavily on Borodkin (1995) and Gerstenblith (2008). We provide only a cursory discussion here; please see these references for further details. 6 In fact, the Department of Homeland Securities guidance to importers makes this point explicitly. See http://www.cbp.gov/linkhandler/cgov/toolbox/legal/informed_compliance_pubs/icp061.ctt/icp061.pdf (downloaded April 2, 2007) for further details.

5

objects there are some limited, albeit important, statutes that do provide some constraints on the

importation of some classes of goods. However, as we discuss below, these constraints are

generally quite limited.

Trade in cultural property is covered in the United States by the Convention on Cultural

Property Implementation Act (CPIA). In the CPIA, cultural property is defined as objects,

collections, specimens, structures, or sites identified as having artistic, historic, scientific,

religious, or social significance.7 The CPIA is the result of the 1970 UNESCO Convention on the

Means of Prohibiting and Preventing the Illicit Import, Export and Transfer of Ownership of

Cultural Property, an international agreement intended to control trade in cultural property. This

agreement required that signatories take steps to make illegal the importation and/or sale of

cultural objects that were removed illegally from any country that was party to the Convention. It

was ratified by the United States in 1972, but required further action by the Congress before its

statutes became law. This was done in a very limited form through the CPIA, resulted in the

implementation of just two sections of the UNESCO convention.

First, the CPIA prohibits the import into the United States of stolen objects that had been

documented in the inventory of a public or secular institution in countries that are signatories to the

Convention. Second, the CPIA grants the President the authority to impose further import

restrictions on specific types of objects through bilateral agreements with other countries. The

other nation must request such an agreement. The United States has signed twelve agreements, and

in eight cases there is variation during 1996-2005.8 However, the existing agreements have mostly

addressed trade only for a narrow range of objects (e.g., Pre-Columbian artifacts from the Petén in

Guatemala and Pre-Classical and Classical archaeological objects in Cyprus). Further, the CPIA

provides only for civil forfeiture of the products in questions and has no criminal penalties. Thus,

overall, the CPIA has very limited coverage and weak punishment.

The U.S. National Stolen Property Act (NSPA) criminalizes the knowing transport, receipt,

and possession of stolen property worth more than 5,000 dollars across international (or state)

7 See Article 1 of the UNESCO Convention for the full definition of cultural property, available at http://www.unesco.org/culture/laws/1970/html_eng/page2.shtml (downloaded on April 17, 2007). 8 We did examine whether the gap between reported exports and imports of antiques and cultural property was affected by the signing of these agreements. Our specifications generally produced coefficients that were consistent with a decreased gap in response to a treaty. However, the results were generally not significant and very sensitive to specification and classification of initial year of treaty and emergency agreements. This is not surprising given the very narrow focus on archaeological objects, the small sample size and the noise in the data. Given the difficulties in interpreting these results, we do not focus on them in this paper.

6

boundaries. The NSPA provides harsher penalties than the CPIA. An individual who knowingly

engages in this conduct can be criminally prosecuted. This effectively allows for the prosecution

for trafficking in objects that can be proven to have come from known collections in a foreign

country. More substantially, the NSPA has been applied in recent years in prosecuting antiquities

smugglers. This derives from laws enacted in most countries with antiquities endowments that

assign ownership of unearthed antiquities to the government. These ownership laws apply to any

objects discovered or excavated after the effective date of the statute. If an object is excavated (or

looted) after this date and removed from the country without permission, then the object is

considered stolen from the government and retains its status as stolen even after it is brought to the

United States. However, we expect this to have a very limited impact on our measure of

smuggling, since antiquities represent only a fraction of the overall of trade in cultural objects.9

Further, it has been difficult in practice for cases to be brought to court under the NSPA: Many

antiquities in the United States were excavated illegally and without the knowledge of the

exporting country’s authorities, and thus lack sufficient documentation to make a case of guilt, and

because the burden of proving that an object has an illegal background falls on the exporting

country’s government or claimant, the proof may be inadequate in a U.S. court to establish that an

undocumented antiquity is an illegally excavated or traded one. If imported cultural objects do not

show up in a U.S. museum (as most do not), the exporter’s government may not be aware of such

trade.

On the other hand, there exists some positive incentive to report truthfully upon entry into

the United States, as improper declaration of the goods upon entry (e.g. lying about the value or

the country of origin) may result in forfeiture. Combined with the zero tariff rate on imports of

cultural objects, there is incentive to report honestly the import of cultural objects.

By contrast, laws on the books in ‘object-rich’ countries generally shift the burden of proof

to the would-be exporter – many nations follow a licensing scheme where permission is required

for export, and others apply their national ownership laws proactively where documentation is

required for export. However, as noted in the introduction, corruption is thought to be rife in many

such countries, and hence exported objects may circumvent legal channels.

9 Personal communication with Professor Patty Gerstenblith. Since trade statistics do not disaggregate antiques into its constituent parts, it is difficult to know the proportion of cultural property and antiques that are antiquities.

7

In summary, there is a stark asymmetry in the reporting imperatives between exporting

nations and the United States. In exporting nations, traders may either not declare cultural objects

to their customs at all or obscure an object’s true value (e.g., label it as a cheap tourist souvenir).

On the importing side, the incentives to provide misleading information are limited, and given the

potential for seizure by the U.S. government for false declaration, there exist some strong

incentives for truthful revelation.

3. Data

Our import and export data come from the World Integrated Trade Solution (WITS) database,

which in turn gets its trade statistics from the United Nations' Comtrade database. These data are

collected by the United Nations Statistical Division from individual countries' trade records, and

include information on imports and exports for each country, recorded according to the 6-digit

Harmonized Commodity Description and Coding System (HS). We use data for all years for which

data are available on imports and exports which results in an unbalanced panel for 1996-2005. We

will also report results below from the balanced panel.

Most export-restricted objects are classified as having HS code 9706 (antiques of an age

exceeding one hundred years).10 Some products in this category are not subject to export controls,

and some products that are subject to restrictions take other classifications. We will also report

results based on an aggregation to the two-digit HS code level (97 – Works of art. Collectors’

pieces, antiques). This is more comprehensive but also potentially incorporates greater noise due to

the inclusion of non-controlled objects. All products in HS code 97 enter the United States tariff

free.

Our primary outcome variable is given by:

(1) Antiques_Gapcy = log(1 + US_Importscy) – log(1 + Exports_to_UScy)

10 See, for example, the European Union guidelines for the protection of cultural property (http://www.culture.gov.uk/NR/rdonlyres/6FC9A8B7-7C91-495F-AC7A-653A45288CC5/0/EUGuidelinesforculturalgoods.pdf downloaded on April 19, 2007).

8

where c indexes country, y indexes year, US_Importscy is the imports reported by the United States

from country c, and Exports_to_UScy is the exports reported by country c destined for the United

States. The industry subscript is suppressed, and is HS Code 9706 unless otherwise noted.

Our primary measure of corruption (Corruptioncy) is from the World Bank Institute

(Kaufman et al, 2006), which generates a composite corruption rating that is essentially the first

principal component of all other available (mostly subjective) corruption indices of country c in

year y. To avoid confusion, we use the negative of the values presented in Kaufman et al so that

values of Corruptioncy are increasing in the level of corruption. This variable is available for 1996,

1998, 2000, 2002-2005. Since virtually all variation is cross-sectional, we use the lagged value of

Corruptioncy for 1997, 1999, and 2001.

It may be useful to account for countries’ endowment of cultural objects, especially those

considered desirable in the major buyer’s markets. Our proxy for this is premised on the

assumption that a country’s endowment of such objects is highly correlated with (or proportional

to) the holdings of that country’s cultural objects by the Metropolitan Museum of Art (Met) in

New York City. The Met’s collection affords a number of advantages in generating a measure for

the potential supply of desirable cultural objects. First, most of its holdings were acquired prior to

the advent of international agreements to control the global flow of cultural property. Second, its

collection is vast, and its mission provides a very general mandate to “collect, preserve, study,

exhibit, and stimulate appreciation for and advance knowledge of works of art that collectively

represent the broadest spectrum of human achievement.” Hence, its collections are not focused on

any particular country or region. An inventory of the Met’s full collection has not yet been put in

digital form. We utilize the listing of the museum’s highlights available on the Met’s webpage,

restricting our attention to pre-19th century non-U.S. collections that would be affected by export

restrictions in the source countries.11

We generate a simple count variable (MetHoldingsc) based on 493 (pre-19th century)

objects listed, reflecting the number of objects in the highlights collection from each country c. In

almost all cases a single country is listed as the object’s origin. Where multiple countries are listed,

we assign partial points equally to all countries. Finally, for 23 objects, ancient regions are listed;

11 http://www.metmuseum.org/Works_of_Art/collection.asp?HomePageLink=permanentcollection_l (downloaded on April 19, 2007). The specific categories that we use are Ancient Near Eastern; Arts of Africa, Oceania & the Americas; Greek & Roman Art; Asian Art; Egyptian Art; Islamic Art; European Paintings; pre-19th century European Sculptures.

9

we assign partial points equally to all countries that overlap geographically with the ancient region.

For example, for the three objects from the Levant, we assign 0.2 points each to Israel, Syria,

Lebanon, Egypt, and Jordan. Given the high dispersion in MetHoldingsc, we will employ an

indicator variable, MetDummyc that denotes whether MetHoldingsc is positive, and also log(1 +

MetHoldingsc) in our analyses.

We use GDP per capita in 2000 U.S. dollars (GDPPCUScy), taken from the World

Development Indicators database, as a control for the overall level of economic development. As

an additional control we use the distance between two countries weighted by the location of their

populations (Distancec), from Mayer and Zignago (2005), which may be reflective of transport

costs. Finally, we will allow for region-year fixed effects, where the regions are North America

and the Caribbean; Latin America; Europe; Africa; Asia; Oceania; and the Middle East.

We restrict our attention for the sample with data available on GDPPCUS cy, Corruption cy,

and Antiques_Gapcy, yielding a final unbalanced sample of 1193 country-year observations for HS

Code 9706 covering 162 countries (the sample will differ slightly when we broaden our sample to

include all trade data for HS Code 97). We present summary statistics in Table 1 for the full

sample, and also the sample split based on countries’ median values of Antiques_Gapcy. Strikingly,

the difference in Corruptioncy between the two groups is 1.14, which is a very large number given

Corruptioncy’s standard deviation of 1.07. However, this may be somewhat confounded by the

correlation with income that is also evident in Table 1. There is also a significant difference in

log(1+MetHoldingsc) and also MetDummyc for the two groups. Finally, we note that there are

many more observations per country for the high Antiques_Gapcy subgroup – this is unsurprising,

as these are countries for which there is a steadier trade in this HS code.

4. Results

We present the raw relationship between Corruptionc and Antiques_Gapcy for the year 2000 in

Figure 1. There is a clear positive relationship (with a correlation coefficient ρ = 0.52) – more

corrupt countries are more likely to under-report in their customs data some of the exports to the

United States relative to the U.S. customs import data. It is interesting to note that many countries

that are well-endowed in ancient objects – Egypt, Syria, Iran, Greece – are well above the

regression line. The reason for this is intuitive – as noted in Section 2, illegal exports constitute

10

only a component of HS 9706, and it is plausible that this component is higher for such countries.

In fact, for MetDummyc=0 countries, the average value of Antiques_Gapcy is 2.90; for

MetDummyc=1 countries, the average value of Antiques_Gapcy is 4.98. This suggests the

importance of controlling for a country’s stock of desirable objects, which we do now in a

regression framework.

Our baseline specification is as follows:

(2) Antiques_Gapcy = α + β1*Corruptioncy + β2*MetDummycy

+ Controlscy + Region-Year Fixed Effects + εcy

While we present results using country-year observations, almost all variation is cross-sectional, so

we allow for clustering at the country level, and also present results using a cross-section with

country-level median values.

We present our main results in Table 2. In column (1) we include only year effects as

controls. As suggested by the pattern in Figure 1, the correlation is highly significant and positive

– the reporting gap in cultural objects is wider for more corrupt countries. In column (2) we

present the results with log of exporter’s income level, log(GDPPCUScy), and year effects only.

There is a negative and significant relationship between the reporting gap and exporter’s income

level, though it is somewhat weaker than that for corruption in column (1). When we include both

income and corruption measures in column (3), however, we find that the point estimate on

Corruptioncy actually increases, while income loses its significance entirely. That is, income

matters only insofar as it is correlated with corruption. Adding Region*Year effects in column (4)

yields very similar results. The magnitude, in the range of 1.5, implies that the rate of smuggling of

cultural objects for relatively high corruption countries such as Mexico or Egypt (Corruptioncy of

about 0.4 in 2005) is more than double that of more moderately corrupt countries such as Italy and

Greece (Corruptioncy of about -0.4 in 2005). This is in part due to some outlying values of

Antiques_Gapcy; however, even when we omit the top and bottom five percent of observations on

Antiques_Gapcy, the significance of the Corruptioncy is largely unchanged and its value is still

above 1.2. We also experimented with specifications that included a variety controls such as

geographic distance, English as primary language, and legal origin; none of these substantively

affected our results and we suppress them to save space.

11

Some of the outlying observations in Figure 1 also suggest the importance of controlling

for countries’ stocks of desirable cultural property – for countries with many desirable objects that

are export restricted, we expect a larger rate of under-reporting for HS 9706 goods. In column (5)

we include MetDummyc and find that it is highly significant and quantitatively large, implying that

Antiques_Gapcy is more than double for countries with objects in the Met’s collection highlights;

the inclusion of MetDummyc has very little effect on the coefficient on Corruptioncy. We also

expect that the marginal impact of corruption to be greater for countries with larger stocks of

cultural objects: In the absence of such objects, the reporting gap should be largely noise, and

hence uncorrelated with corruption; we expect a larger effect of corruption as the potential for

smuggling increases. We report these results in column (6) and find that the interaction term is

highly significant with a magnitude of about 0.5.12 In columns (7) and (8) we repeat our analyses

using log(1+MetHoldingsc) in place of MetDummyc; the implied magnitudes are very similar for

both measures.

We present results for four additional countries that are potential destination markets for

cultural objects. Three of these – Switzerland, Great Britain, and Germany – are noteworthy in that

none ratified the 1970 UNESCO Convention until very recently (Germany is still not a signatory to

the Convention). By contrast, our fourth country – Canada – ratified the UNESCO Convention and

passed the Cultural Property Export and Import Act in 1978 that provides stringent controls on

importing cultural objects that were illegally exported. However, by many accounts enforcement

has been spotty, and only five cases of illegal imports have been resolved since 1992 (Department

of Canadian Heritage, 2003, 2005). We present results using our preferred specification that

includes Region*Year dummies in Table 3. In all cases, the coefficient on Corruptioncy is

significant at the one percent level. In all cases, the coefficient is in the range of 0.5 – 1.13 This is

both surprising and interesting given the range of legal statutes across the four markets:

Switzerland has a reputation as a haven for laundering the provenance of ancient art, whereas

Canada has potentially strong legal sanction against trafficking in cultural property.

Finally, in Table 4 we present a range of robustness tests. In Column (1) we report results

using exporter-level medians to take away the time-series element of our variation. The coefficient

12 If we include log(GDPPCUScy)*MetDummyc as a control, we find it to be insignificant, and the coefficient on Corruptoncy*MetDummyc is unaffected. 13 Perhaps surprisingly, the coefficient on MetDummyc does not take on any consistent sign across specifications. This may reflect different tastes for artifacts across countries, as our measure is U.S.-based.

12

on Corruptioncy is 1.30, marginally lower than the coefficients reported in Table 2. Column (2)

presents results for imports and exports for HS Code 97 (Works of art. Collectors’ pieces,

antiques), since some cultural property may be classified in other 4-digit classes, such as paintings

(9701) or sculpture (9703). We obtain similar, though somewhat weaker, results than those

reported in Table 2. As previously noted, this may result from the fact that products in HS 97 but

outside of HS 9706 would be expected to contain a smaller proportion of export-restricted objects.

In Column (3) we show our results with exporting country fixed effects included. While the

coefficient on Corruptioncy is still positive, it is no longer significant. This is not surprising, given

that most of the variation in Corruptioncy is cross-sectional. Finally, we include results using HS

Code 9503 (Toys, scale models etc, puzzles, parts), by far the largest component of the 2-digit HS

Code 95; we also report results using the entire HS 95 industry. As a placebo regression this

industry has the advantages of having had a zero tariff rate since 1994 (and hence no incentive for

importers to lie to the US customs). We report results for HS 9503 and HS 95 in columns (4) and

(5) respectively. In both cases, the coefficient on Corruptioncy is indistinguishable from zero. This

suggests that the positive association between exporter’s corruption and the reporting gap for the

imports of cultural objects to the U.S. is unlikely due to some missing factors that are common

across product lines.

5. Conclusions

Exploring different reporting incentives in the trade in cultural objects and antiques between the

exporter (e.g., Egypt) and the importer (e.g., the United States) sides, this paper provides a gauge

for illicit trade in cultural goods. We find strong and robust evidence that the percentage under-

recording of exports of cultural objects is highly correlated the exporting country’s level of

corruption as measured by a commonly used subjective index. Furthermore, the association

between the two is stronger for countries that are particularly well-endowed in export-restricted

objects that are considered to be desirable in the major markets.

We provide a number of contributions. First, we present a simple methodology that can be

applied to generate cross-country estimates of illicit trade. This has become feasible only in recent

years when large and highly-disaggregated trade data from both importing and exporting countries

13

have been made available.14 While some additional creativity will be required to apply this method

to other types of illicit trade, we believe that the approach will prove to be useful. For example,

some legal inputs are required in the production of illegal drugs: potassium permanganate is used

to produce crack cocaine, but also many other products. As a result, it is a controlled substance in

some countries, such as Colombia, but much less tightly regulated in others. This type of legal

asymmetry may similarly lead to different reporting incentives which could be utilized in tracking

other illicit trade.

Second, we provide an important contribution to the literature on measuring corruption.

Since mid-1990s, there has been an explosion in the use of corruption indices in empirical

research.. Because corruption is illegal in most countries, almost all available measures are

subjective indices based on surveys of citizens, experts, or firms. By finding a clear association

between smuggling in cultural objects based on objectively collected trade data and a commonly

used subjective corruption ranking, this paper provides valuable and independent confirmation that

the survey-based corruption indices contain useful information.

While it is tempting to try to use our results to calibrate the total volume of trafficking in

cultural objects into the United States and worldwide, it is not possible to generate any meaningful

measure based on our analyses. First, the ‘trade gap’ between reported imports and reported

exports is generally positive for all goods – we are interested in the correlates of this gap rather

than the level of the gap itself. Further, if we wish to use our regression results for such

calculations, the numbers we produce will be highly sensitive to our assumptions of the extent of

trafficking from very low corruption source countries. Given that our results are expressed in terms

of elasticities, any change in this assumption will naturally generate a proportionate increase in our

final measure of the total level of trafficking. We will leave this type of exercise for future work.

14 As a practical implication for law enforcement communities, real-time cross-checking of export and import declarations may provide an extra tool to capture smuggling.

14

References

Andreas, Peter (1998). “Transnational Crime: Law Enforcement and Law Evasion in a Changing

World,” Transnational Organized Crime, pp. 97-90.

Atwood, Roger, Stealing History: Tomb Raiders, Smugglers, and the Looting of the Ancient

World, New York: St. Martin Press, 2004, pg. 221.

Borodkin, Lisa (1995). “The Economics of Antiquities Smuggling and a Proposed Legal

Alternative.” Columbia Law Review, March 1995, pp. 377-417.

Conklin, John E (1994). Art Crime. Westport, CT: Praeger.

Brodie, Neil, Jenny Doole and Peter Watson, “Stealing History: The Illicit Trade in Cultural

Material,” McDonald Institute for Archaeological Research, 2000.

Fisman, Raymond, and Edward Miguel (2007). “Corruption, Norms and Legal Enforcement:

Evidence from Diplomatic Parking Tickets,” unpublished manuscript.

Fisman, Raymond, and Shang-Jin Wei (2004). “Tax Rates and Tax Evasion,” Journal of Political

Economy, vol. 112, pp. 471–496.

Mayer, Thierry and Soledad Zignago, “Market Access in Global and Regional Trade,” CEPII

Working Paper 2005-02, 2005.

Mishra, Prachi, Petia Topalova, and Arvind Subramanian (2007). “Policies, Enforcement, and

Customs Evasion: Evidence from India,” IMF Working Paper No 07/60.

Prott, Lyndell, and Patrick O’Keefe (1988). Handbook of National Regulations Concerning the

Export of Cultural Property, UNESCO.

15

Yasaitis, Kelly Elizabeth. “National Ownership Laws as Cultural Property Protection Policy: The

Emerging Trend in United States v. Schultz,” International Journal of Cultural Property (2005)

12:95–113. (Reference for national ownership law as primary means of enforcing things)

Yang, Dean (2007). “Can Enforcement Backfire? Crime Displacement in the Context of Customs

Reform in the Philippines.”, forthcoming, Review of Economics and Statistics.

Toner, Michael (1999). "The Past in Peril." Atlanta Journal and Constitution, September 19, 1999.

16

Figure 1: Percentage Under-reporting of Exports of Cultural Objects

to the US and Exporters’ Corruption Level, 2000

AFG

AGO

ALBARE

ARG

ARM

AUSAUT

AZE

BEL

BEN

BFABGD

BGR

BHSBIH

BOL

BRA

BTN

CAF

CAN

CHE

CHL

CHNCIV

CMR

COG

COK

COL

CRI

CYP

CZEDEU

DNK

DOMDZA

ECU

EGY

ESP

EST

ETH

FIN

FJI

FRA

GAB

GBR

GEO

GHA

GIN

GNB

GRC

GTM

HKG

HND

HRV

HTI

HUN

IDN

IND

IRL

IRN

ISLISR

ITA

JAM

JOR

JPN

KAZ

KEN

KGZ

KHM

KOR

LBN

LBR

LKA

LTU LVA

MAR

MDGMEX

MKD

MLI

MMR

MNGMUS

MYS

NAM

NGA

NLD

NOR

NPL

NZL

PAK

PAN

PER

PHL

PNG

POLPRT

PRY

ROM

RUS

SAU

SGP

SLB

SLESUR

SVK

SVN

SWE SWZ

SYC

SYR

TGO

THA

TKM

TON

TUN

TUR

TWN

TZAUGA

UKR

URY

UZB

VEN

VNM

YEM

ZAF

ZAR

-50

510

-3 -2 -1 0 1 2(mean) corrupt

antiques_gap Fitted values

17

Table 1 - Summary Statistics Mean Std. Dev. Min Max Obs Antiques_Gap 3.83 2.70 -3.58 10.65 1199 Corruption -0.07 1.07 -2.52 2.13 1193 log(GDPPCUS) 7.82 1.60 4.45 11.21 1193 MetDummy 0.40 0.49 0.00 1.00 162 log(1+MetHoldings) 0.59 0.95 0.00 4.04 162 log(Distance) 9.07 0.44 7.64 9.71 161 Above median Antiques_Gap Corruption 0.49 0.57 -1.72 2.13 602 log(GDPPCUS) 7.11 1.18 4.45 10.09 602 MetDummy 0.62 0.49 0.00 1.00 69 log(1+MetHoldings) 0.91 1.05 0.00 4.04 69 log(Distance) 9.13 0.42 7.92 9.65 68 Below median Antiques_Gap Corruption -0.65 1.15 -2.52 1.36 591 log(GDPPCUS) 8.54 1.66 4.95 11.21 591 MetDummy 0.24 0.43 0.00 1.00 93 log(1+MetHoldings) 0.35 0.79 0.00 3.78 93 log(Distance) 9.03 0.46 7.64 9.71 93 Notes: Antiques_Gap is defined as log(1+US_Imports) - log(1+Exports_to_US), where US_Imports are imports reported by the United States for HS Code 9706 from country c, and Exports_to_US are exports destined for the United States for HS Code 9706 from country c. Corruption is (the negative of) the Kaufman et al (2006) measure of corruption. GDPPCUS is per capita income for country c in constant 2000 US dollars. MetDummy denotes that country c is reported as the country of origin for at least one item in the Metropolitan Museum of Art's highlights collection. MetHoldings is the number of pre-19th Century items in the Metropolitan Museum's highlights collection. Corruption and GDPPCUS are country-year level observations; MetDummy and MetHoldings are country-level observations. Please see section 2 of the text for further information and sources.

18

Table 2 - Correlation between corruption and smuggling of cultural objects (1) (2) (3) (4) (5) (6) (7) (8) Corruption 1.371*** 1.602*** 1.468*** 1.404*** 1.165*** 1.385*** 1.117*** (0.111) (0.222) (0.235) (0.213) (0.219) (0.211) (0.192) log(GDPPCUS) -0.720*** 0.182 0.014 0.005 0.004 -0.024 0.027 (0.098) (0.163) (0.199) (0.176) (0.175) (0.174) (0.162) MetDummy 1.105*** 1.153*** (0.280) (0.272) MetDummy 0.535*** *Corruption (0.202) log(1 + MetHoldings) 0.783*** 0.930*** (0.161) (0.131) log(1 + MetHoldings) 0.519*** *Corruption (0.103) Fixed Effects Year Year Year Region*Yr Region*Yr Region*Yr Region*Yr Region*YrObservations 1193 1193 1193 1193 1193 1193 1193 1193 R-squared 0.30 0.18 0.30 0.40 0.75 0.75 0.75 0.75 Notes: Robust standard errors in parentheses, clustered by country. The dependent variable in all regressions is Antiques_Gap, which is defined as log(1+US_Imports) - log(1+Exports_to_US), where US_Imports are imports reported by the United States for HS Code 9706 from country c, and Exports_to_US are exports destined for the United States for HS Code 9706 from country c. Corruption is (the negative of) the Kaufman et al (2006) measure of corruption. GDPPCUS is per capita income for country c in constant 2000 US dollars. MetDummy denotes that country c is reported as the country of origin for at least one item in the Metropolitan Museum of Art's highlights collection. MetHoldings is the number of pre-19th Century items in the Metropolitan Museum's highlights collection. Corruption and GDPPCUS are country-year level observations; MetDummy and MetHoldings are country-level observations. Please see section 2 of the text for further information and sources. * significant at 10%; ** significant at 5%; *** significant at 1%

19

Table 3 - Predicting the Antiques_Gap in other importing countries (1) (2) (3) (4) Corruption 0.582** 0.807*** 0.761** 0.924*** (0.286) (0.248) (0.335) (0.323) log(GDPPCUS) -0.134 0.035 -0.073 0.241 (0.231) (0.205) (0.241) (0.256) MetDummy -0.222 0.996*** -0.030 0.274 (0.328) (0.323) (0.348) (0.325) Importing Country Switzerland Germany Great Britain Canada Observations 470 483 692 528 R-squared 0.38 0.35 0.33 0.27 Notes: Robust standard errors in parentheses, clustered by country. The dependent variable in all regressions is Antiques_Gap, which is defined as log(1+Imports_by_k) - log(1+Exports_to_k), where Imports_by_k are imports reported by country k (Switzerland, Germany, UK or Canada) for HS Code 9706 from country c, and Exports_to_k are exports destined for country k for HS Code 9706 as reported by country c. Corruption is (the negative of) the Kaufman et al (2006) measure of corruption. GDPPCUS is per capita income for country c in constant 2000 US dollars. MetDummy denotes that country c is reported as the country of origin for at least one item in the Metropolitan Museum of Art's highlights collection. Corruption and GDPPCUS are country-year level observations; MetDummy is a country-level observations. Please see section 2 of the text for further information and sources. * significant at 10%; ** significant at 5%; *** significant at 1%

20

Table 4 - Robustness Tests (1) (2) (3) (4) (5) Corruption 1.298*** 0.983*** 0.291 -0.090 -0.036 (0.198) (0.205) (0.311) (0.197) (0.203) log(GDPPCUS) 0.021 -0.154 0.305 -0.281** -0.553*** (0.143) (0.151) (0.390) (0.139) (0.175) MetDummy 1.396*** 1.098*** 0.091 0.099 (0.288) (0.244) (0.209) (0.308) HS Code 9706 97 9706 9503 95 Fixed Effects Region Region*Year Country & Yr Region*Year Region*Year Observations 162 1439 1193 972 1221 R-squared 0.50 0.34 0.75 0.19 0.19 Notes: Robust standard errors in parentheses, clustered by country. The dependent variable in all regressions is Antiques_Gap, which is defined as log(1+US_Imports) - log(1+Exports_to_US), where US_Imports are imports reported by the United States from country c, and Exports_to_US are exports destined for the United States from country c. The relevant industry code is listed above in the Table. Specification (1) employs a cross-section with country-level medians from 1996-2005 of all variables; in (2) - (5) uses all years individually. Corruption is (the negative of) the Kaufman et al (2006) measure of corruption. GDPPCUS is per capita income for country c in constant 2000 US dollars. MetDummy denotes that country c is reported as the country of origin for at least one item in the Metropolitan Museum of Art's highlights collection. Corruption and GDPPCUS are country-year level observations; MetDummy is a country-level observations. HS Code 9706 is Antiques of an age exceeding one hundred years; HS Code 97 is Works of art, Collectors’ pieces, antiques. HS 9503 is Toys, scale models etc, puzzles, parts. HS Code 95 is Toys, games and sports requisites; parts and accessories thereof. Please see section 2 of the text for further information and sources. * significant at 10%; ** significant at 5%; *** significant at 1%